Green Index Calculation for OpenRouteService

This repository contains the source code to

1. calculate the greenness (i.e. presence of vegetation) based on OpenStreetMap (OSM) and Sentinel-2 data [1]
2. calculate the green index of each OSM highway feature to be used openrouteservice to calculate green routes.
3. compare short and green routes using openrouteservice.

Using this code the analysis presented at the GIScience Conference 2021 [2] can be applied to other geographic regions. Note that for this to be possible the code had to be adapted slightly, so the results from the paper cannot be reproduced exactly. Also note that although the green index can be calculated for any region, the quality of the resulting green index and the green routes depends on the quality of the underlying OSM and Sentinel-2 data.

Installation

Python 3 and the packages listed in requirements.txt are required. You can set up a new python environment with all dependencies using pip:

$ python3 -m venv env
$ source env/bin/activate
$ python3 -m pip install -r requirements.txt
$ pip install git+git://github.com/reineking/pyds.git

1. Greenness calculation

The first step is to calculate the greenness of individual street blocks within the area of interest using the calculate_greenness.py script.

Example:

$ greenness.py -c config_sample.json -g google_credentials_sample.json

Usage:

$ python greenness.py -h

usage: greenness.py [-h] --config CONFIG_FILE --google_cred
                              GOOGLE_CRED_FILE

Calculates the greenness based on OSM and Sentinel-2 data.

optional arguments:
  -h, --help            show this help message and exit
  --config CONFIG_FILE, -c CONFIG_FILE
                        Path to configuration file (.json) describing the area of interest (aoi)
  --google_cred GOOGLE_CRED_FILE, -g GOOGLE_CRED_FILE
                        Path to json file containing credentials for using
                        google earth engine

Configuration file: -c / --config

The configuration file is a json file which contains all parameters required to describe the area of interest, e.g. ./config/config_sample.json

{
  "name": "sample",
  "bbox": [8.46874,49.4971,8.49213,49.50995],
  "epsg": "32632",
  "timestamp": "2021-08-15",
  "cloud_coverage": 5,
  "ndvi_year": 2020,
  "output_dir": "./data"
  "fuzzy_centers": {
    "green": 0.71,
    "mixed": 0.43,
    "grey": 0.15,
    "d": 0.094
  }
}

Parameter	Description
name	The name of the area of interest. Used in output file names.
bbox	Bounding box of area of interest in geographic coordinates, format: (minx, miny, max, maxy)
epsg	The epsg of a projected coordinate reference system suitable for the are of interest.
timestamp	Timestamp of the OSM data used for processing.
ndvi_year	Year used to calculate the annual maximum NDVI.
output_dir	Path to a existing directory in which output data will be stored.
fuzzy centers	NDVI values used as centres to distinguish green from non-green areas. The default parameters are adapted to Dresden, Germany.

Google Credentials File: -g / --google_cred

The calculation of the NDVI is done using Google Earth Engine. To use this service you need to create create a service account with google and generate a key. Put these in the ./config/google_credentials_sample.json

{
  "service_account": "XXXX@XXXX.iam.gserviceaccount.com",
  "service_account_json": "./config/quickstart-XXX-XXX.json"
}

2. Green Index Calculation

After the greenness is calculated, the green index of each OSM highway feature can be calculated. This script will

• download the OSM highway features using the ohsome API and
• calculate the green index of each highway feature.

Example:

$ python green_index.py -b 13.7081,51.0372,13.7747,51.0678 -r ./data/sample/sample_greenness.tif -w 20 -o ./data/sample

Usage:

$ python green_index.py -h

Usage: green_index.py [-h] --bbox BBOX [--timestamp TIMESTAMP] --width WIDTH [--vector VECTOR_FILE] [--raster RASTER_FILE] --out_dir OUTPUT_DIR

Calculates the index of each OSM highway based on provided raster or vector file. The highways are downloaded using the ohsome API.

optional arguments:
  -h, --help            show this help message and exit
  --bbox BBOX, -b BBOX  Bounding box in geographic coordinates as string without whitespace e.g. 'minx,miny,maxx,maxy')
  --timestamp TIMESTAMP, -t TIMESTAMP
                        ISO formatted timestamp for download of highways from OSM. By default the latest timestamp available in the ohsome API will be used.
  --width WIDTH, -w WIDTH
                        Width of the buffer around highway segments in meters
  --vector VECTOR_FILE, -v VECTOR_FILE
                        Path to vector file containing features counted nearby highways
  --raster RASTER_FILE, -r RASTER_FILE
                        Path to raster file used to calculate mean value within area nearby highway
  --out_dir OUTPUT_DIR, -o OUTPUT_DIR
                        Path to output directory.

3. Comparison of green and short routes

3.1 Set up openrouteservice instance

Last step is to set up an instance of the openrouteservice with the produced green index.

1. Fork and clone this this openrouteservice repository. Checkout out the branch green_index. In this branch, the green index of the original openrouteservice code has been slightly adapted.
2. Download the OSM data for your area of interest from Geofabrik.
3. Put the green index csv file and an OSM file in the folder openrouteservice/docker/data.
4. Replace the files in the openrouteservice/docker/docker-compose.yml to match your files.

3.2. Comparison of routes

As a last step random foot or bike trips can be simulated and the respective green and short routs are compared.

Example:

$ python route_comparison.py -c ./config_comparison_sample.json

Usage:

$ python route_comparison.py -h

usage: route_comparison.py [-h] --config CONFIG_FILE

Comparison of short and green routes using openrouteservice

optional arguments:
  -h, --help            show this help message and exit
  --config CONFIG_FILE, -c CONFIG_FILE
                        Configuration file

{
  "name": "sample",
  "out_dir": "./data",
  "districts_file": "./data/districts/districts_sample.geojson",
  "col_name": "bez",
  "epsg": "32633",
  "profile": "foot-walking",
  "steep_level": 1,
  "n_routes": 100,
  "min_length" : 200,
  "max_length" : 5000,
  "ors_url": "http://localhost:81/ors/"
}

Parameter	Description
name	The name of the area of interest. Used in output file names.
out_dir	Path to a existing directory in which output data will be stored.
districts_file	Vector file containing the districts of the region
col_name	Column name to use a unique district name.
epsg	The epsg of a projected coordinate reference system suitable for the are of interest.
profile	"foot-walking" for pedestrian or "cycling-regular" for bike routes.
steep_level	Value between 0 (avoid slopes) and 3 (do not avoid slopes)
n_routes	Number of routes to be generated for each pair of districts
min_length	Minimum length of route
max_length	Maximum length of route
ors_url	URL to local openrouteservice instance

Related publications

[1]Ludwig, C., Hecht, R., Lautenbach, S., Schorcht, M., & Zipf, A. (2021). Mapping Public Urban Green Spaces Based on OpenStreetMap and Sentinel-2 Imagery Using Belief Functions. ISPRS International Journal of Geo-Information, 10(4), 251. doi: https://doi.org/10.3390/ijgi10040251

[2] Ludwig, C., Lautenbach, S., Schömann, E., Zipf, A. (2021): Comparison of fast and green routes for cyclists and pedestrians. GIScience Conference 2021.